Attitudes, Incentives, and Tax Compliance

(2005) vol. 53, no 1 ■ 29

Attitudes, Incentives, and Tax Compliance

Viswanath Umashanker Trivedi, Mohamed Shehata, andStuart Mestelman*

P R É C I S

Des données publiées par l’Agence du revenu du Canada montrent que l’observationfiscale à son niveau le plus simple (production et paiement à temps) est plutôtsatisfaisante tandis qu’elle devient plus problématique à des niveaux complexes, si onse fie à la proportion de contribuables soumis à des risques d’inobservation plusimportants. Pour savoir si une augmentation du nombre de vérifications et du montantdes pénalités constitue la meilleure façon de contrer l’inobservation, il faut connaîtreles raisons pour lesquelles les contribuables font défaut de se conformer. Si lescontribuables ne s’intéressent qu’aux allégements et s’en remettent au hasard de la« loterie de la vérification », l’augmentation du montant des taux des pénalités et dunombre de vérifications devrait contribuer à améliorer l’observation. Mais si lesfacteurs psychologiques (préoccupations d’ordre moral et éthique) sont importants, onpourrait encourager l’observation par l’élaboration de stratégies qui modifientl’attitude des contribuables envers le régime fiscal, par exemple en améliorant laperception de son caractère équitable et en facilitant la conformité aux lois fiscales.

Cette analyse s’ajoute à la doctrine sur les mesures appropriées d’observationfiscale à utiliser pour déterminer quels changements aux lois fiscales encourageraientl’observation. Dans le passé, certaines études utilisaient des mesures fondées sur desquestions d’observation fiscale dans des situations hypothétiques tandis que d’autresétudes avaient recours à des mesures fondées sur des méthodes économiquesexpérimentales. Cet article conclut que ce sont ces méthodes qui produisent la mesurequi offre la meilleure corrélation avec la participation volontaire à l’évasion fiscale, dumoins si les méthodes sont du type décrit ci-après où une terminologie propre à lafiscalité est utilisée.

Un deuxième objectif de cet article est d’utiliser une expérience de contrôle quiévalue l’observation fiscale chez les sujets (des étudiants) de l’étude pour vérifier :

* Viswanath Umashanker Trivedi is of the Schulich School of Business, York University, MohamedShehata of the MGD School of Business, McMaster University, and Stuart Mestelman of theDepartment of Economics, McMaster University. We thank participants in the Tax PolicyResearch Symposium: Perspectives from Law and Accounting held in August 2004 in Torontofor helpful comments. This article has also significantly benefited from the comments andrecommendations of the editor, Alan Macnaughton, and two anonymous reviewers. Thisresearch was funded by McMaster University Arts Research Board grants to Mohamed Shehataand Viswanath Umashanker Trivedi.

30 ■ canadian tax journal / revue fiscale canadienne (2005) vol. 53, no 1

1) les prédictions de la théorie économique et 2) une théorie psychologique particulièresur la raison pour laquelle des contribuables pourraient se conformer ou non à la loi(théorie de Ajzen sur le comportement planifié). Dans la tradition des méthodeséconomiques expérimentales, les participants à l’expérience se voient accorder desincitatifs monétaires réels pour motiver leur comportement. Cependant, pour introduiredes facteurs psychologiques, tels les comportements, dans l’aspect moral de l’évasionfiscale, cette étude se démarque de la plupart des recherches économiquesexpérimentales passées par l’utilisation d’une terminologie propre à la fiscalité pourdécrire aux participants le processus de prise de décision. Les résultats confirmentgénéralement les prédictions de la théorie de Ajzen et soulignent l’importance de tenircompte des attitudes et des intentions de la personne au sujet de l’observation fiscaleen plus des considérations d’ordre économique. D’autres recherches seront nécessairespour confirmer nos conclusions et déterminer la meilleure façon de les appliquer dansun contexte de modalités avant de les utiliser dans un contexte décisionnel.

A B S T R A C T

Data published by the Canada Revenue Agency show that tax compliance at the mostbasic level (filing and remitting on time) is quite satisfactory, while the more significantlevels of tax compliance, as measured by the proportion of taxpayers judged to be at“substantive risk of non-compliance,” show more problems. Whether an increase inaudits and penalties is the best way to deal with this non-compliance depends on thereasons why taxpayers fail to comply. If taxpayers care only about incentives and are“playing the audit lottery,” increasing penalty and audit rates should improvecompliance. But if psychological factors (including moral and ethical concerns) areimportant, improved compliance might instead be achieved by strategies that changetaxpayers’ attitudes toward the tax system, such as increasing its perceived fairnessand making it easier to comply with the tax laws.

This study contributes to the compliance literature by examining the appropriatemeasure of tax compliance to use in research aimed at determining those changes intax laws that would encourage higher levels of compliance. In the past, some studieshave used measures based on questions about tax compliance in hypothetical situations,while others have used measures based on the experimental-economics methodology.This paper finds that experimental-economics methods produce the measure that has abetter correlation with self-reported participation in tax evasion, at least if theexperimental-economics work is of the type described below in which tax-specificterminology is used.

A second purpose of the paper is to use a controlled experiment that assesses taxcompliance among student subjects to test (1) predictions of economic theory and (2) aparticular psychological theory of why taxpayers might comply or fail to comply (Ajzen’stheory of planned behaviour). In the tradition of experimental economics, participantsin the experiment are given real monetary incentives to motivate their behaviour.However, in order to introduce psychological factors such as attitudes about the moralityof tax evasion, this study breaks with most past experimental-economics work by usingtax-specific terminology to describe the decision-making situation to participants. Theresults generally confirm the predictions of Ajzen’s theory and highlight the importanceof considering individuals’ attitudes and intentions toward tax compliance over andabove pure economic considerations. Subsequent research to confirm our findings and

attitudes, incentives, and tax compliance ■ 31

to determine how best to apply them in policy contexts is needed before such findingscan be applied to policy decisions.

KEYWORDS: COMPLIANCE ■ TAX EVASION ■ TAXPAYERS ■ ECONOMICS ■ MODELS ■ EXPERIMENTS ■

ETHICS

C O N T E N T S

Introduction 31Encouraging Tax Compliance 31Alternative Research Approaches 37Purposes of This Study 38

Measuring Tax Compliance Behaviour 40Overview of the Survey and Experiment 40Compliance Measures 41Results 44

Levels of Compliance 44Correlations Among Compliance Measures 46Compliance by Situation 47

Testing a Theory of Tax Compliance 48Variables Suggested by the Theory of Planned Behaviour 48

Subjective Norms Factor 50Attitudes Factor 50Perceived Behavioural Control Factor 51

Tax Compliance Intentions 51Ethics 51Results 53

Tests of Hypothesis 1 54Tests of Hypothesis 2 58

Conclusion 58

I N T R O D U C T I O N

Encouraging Tax Compliance

The Canada Revenue Agency’s (CRA) assessment of the current level of compliancewith Canadian tax laws is that “while non-compliance overall is material, it remainsrelatively low, in line with prior years and comparable to other OECD [Organisationfor Economic Co-operation and Development] countries.”1 Still, non-complianceis not a problem that can be safely ignored. The data in table 1 show that whilecompliance at the most basic level (filing and remitting on time) is quite satisfactory,the more significant levels of compliance (as measured by the proportion of taxpay-ers judged to be at “substantive risk of non-compliance”) exhibit more problems.In particular, only 8 percent of individuals do not file on time and only 6 percent do

1 Canada Customs and Revenue Agency, CCRA Annual Report to Parliament 2003-2004 (Ottawa:Canada Customs and Revenue Agency, 2004), 1-24, exhibit 6.


not remit on time; rates for taxable corporations are higher but still low at 13 percentand 9 percent, respectively. In contrast, 29 percent of self-employed individuals and38 percent of corporations are believed by the CRA to be at substantive risk of non-compliance. Although this information is subjective and cannot be independentlyverified, it reflects the only data available that are drawn from actual taxpayer recordsin Canada.2

The CRA, of course, attempts to reduce non-compliance. Its attempts appear tobe effective as judged by its fiscal impact data, which the auditor general has judgedto be “a fair and reliable account, in most significant respects.”3 The CRA’s estimateis that the net fiscal impact of its compliance activities is a contribution of approxi-mately $8 billion to the public purse (see panel D, table 1), which is up 70 percentin the three years from 2000-1 to 2003-4.4 This rising trend is perhaps somewhatsurprising, in that audit rates appear to have been relatively flat over the sameperiod5 and penalty rates have not increased.

2 Much more data are made public by the United States Internal Revenue Service (IRS). See ValentinEstevez, Liberals, Conservatives and Your Tax Return: Partisan Politics and the Enforcement Activities of theIRS, Job Market Paper (Chicago: University of Chicago, Department of Economics, 2004) (available onthe Web at http://home.uchicago.edu/~vestevez/jobmarketpaper.pdf ); and United States, InternalRevenue Service, IRS Databook (Washington, DC: Internal Revenue Service, various editions).

3 Canada Customs and Revenue Agency, CCRA Annual Report to Parliament 2002-2003 (Ottawa:Canada Customs and Revenue Agency, 2003), 1-97.

4 The CRA spent $709 million dollars on compliance programs during the 2003-2004 period(supra note 1, at 1-22, exhibit 4). Thus, as a crude inference, it appears that the CRA collectsapproximately $11 from its compliance activities for every $1 spent. Another figure to compareto the $8 billion is the total amount of $313 billion dollars (including customs deposits) thatthe CRA deposited into the Consolidated Revenue Fund during 2003-2004 (ibid.) It thusappears that something like 97 percent of the funds collected are paid without the requirementof CRA intervention. A discussion of whether these numbers are high or low is outside thescope of this paper.

5 It seems reasonably safe to conclude from CRA statistics (see table 1) that audit rates for thelargest corporations have remained fairly constant in recent years at about 55 to 60 percent, whilesmall business audit rates have increased from about 0.8 percent to about 1.2 percent. However,the audit rate for individuals is a murky concept, and trends in this rate are even harder tointerpret. Table 1 appears to show that the audit rate for individuals has remained fairly constantat no more than 1.3 percent; but this rate is far too low, because it does not include two majorcompliance verification programs directed at non-business individuals. First, the matching oftax slips is used to verify the returns of the vast majority of tax filers; over 1 million taxpayerswere reassessed for this reason in 2003-2004. Second, there is the validation program (forexample, requests for charitable slips and medical expense slips for e-filers), which reviewed thedeductions and credits of over 744,000 tax returns in 2003-2004. However, it is unclear howeffective the CRA is in (1) identifying non-compliant taxpayers and (2) in unearthing the fullextent of the non-compliance of those identified as being non-compliant. In this regard,evidence from the Taxpayer Compliance Measurement Program (TCMP) in the United Statessuggests that as much as half the amount of evasion remains undetected even under therigorous TCMP audits. See Jonathan S. Feinstein, “An Econometric Analysis of Income TaxEvasion and Its Detection” (1991) vol. 22, no. 1 RAND Journal of Economics 14-35. InternalRevenue Service, Income tax compliance research: Supporting appendices to publication 7285,publication 1415 (Washington, DC: IRS, July 1988), A-31.


TABLE 1 CRA Disclosures Regarding Taxpayer Compliancea

Panel A: Audit coverage rates relating to audit type activity

2005-2000- 2001- 2002- 2003- 20062001 2002 2003 2004 target

percentLarge businesses

Basic files . . . . . . . . . . . . . . . . . . . . . . . . . 12.00b 13.00c 21.80c 25.00c 20.80d

Large files . . . . . . . . . . . . . . . . . . . . . . . . 54.00b 54.20c 56.60c 56.60c 60.00d

Small businessesUnincorporated . . . . . . . . . . . . . . . . . . . . 0.68b 0.82c 1.01c 0.97c 1.16d

Corporate . . . . . . . . . . . . . . . . . . . . . . . . . 0.87b 0.95c 0.99c 0.90c 1.26d

GST/HST post-payment audits . . . . . . . . . 0.80b 0.82c 1.04c 0.96c 1.38d

numberAudits, examinations, and reviews

completed (A) . . . . . . . . . . . . . . . . . . . . . . . 288,107f 282,974g 301,714h 304,707i naj

Other enforcement actionse . . . . . . . . . . . . . 2,663f 3,093g 3,990h 3,327i na

numberFiles processed (in millions)

Individual (B) . . . . . . . . . . . . . . . . . . . . . . 22.80f 22.80g 22.90h 22.70i naCorporate . . . . . . . . . . . . . . . . . . . . . . . . . 1.30f 1.60g 1.40h 1.50i naGST/HST . . . . . . . . . . . . . . . . . . . . . . . . 6.10k 6.40k 6.80h 6.90i na

percentAudit rate for individuals—upper limit:

(A)/(B) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.26 1.24 1.32 1.34 na

Panel B: Estimated percentage of business accounts deemedto be at substantive risk of non-compliance

2000- 2001- 2002- 2003-2001l 2002m 2003m 2004m

percent

GST/HST accounts . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.00 17.00 18.00 17.00Individuals/self-employed . . . . . . . . . . . . . . . . . . . . . . . 19.00 22.00 27.00 29.00Corporate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29.00 31.00 38.00 38.00

(Table 1 is continued on the next page.)


TABLE 1 Continued

Panel C: Estimated non-compliance rates

2000- 2001- 2002- 2003-2001 2002n 2003o 2004o

percentFiling non-compliance (% who did not file on time)

Individuals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.40p 6.10 5.70 7.90Taxable corporations . . . . . . . . . . . . . . . . . . . . . . . . 8.20p 11.10 12.80 12.90GST/HST returns . . . . . . . . . . . . . . . . . . . . . . . . . 10.30p 9.70 8.40 8.50Employers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . na 3.60 3.60 3.50

GST/HST registration (% not registeredwhere required) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23.70p 18.80 15.40 13.40

Remittance non-compliance (% who did notpay tax on time)

Individuals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.60q 9.70 6.90 5.70Taxable corporations . . . . . . . . . . . . . . . . . . . . . . . . 8.00q 5.00 7.00 9.00Employers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . na 10.10 9.90 9.90

Panel D: Total gross fiscal impact of tax complianceactivities (audit and enforcement activities)

2000- 2001- 2002- 2003-2001r 2002s 2003t 2004u

Large business audit, $ millions . . . . . . . . . 2,066 2,017 2,417 2,185(34.45%) (31.10%) (26.18%) (21.42%)

Small and medium enterprise audits,$ millions . . . . . . . . . . . . . . . . . . . . . . . . . 1,403 1,430 1,617 1,637

(23.40%) (22.05%) (17.51%) (16.05%)Non-filersv, $ millions . . . . . . . . . . . . . . . . . 750 1,418 1,655 2,046

(12.51%) (21.87%) (17.92%) (20.06%)GST/HST and employer compliance,

$ millions . . . . . . . . . . . . . . . . . . . . . . . . . 2,027 2,340na na (21.95%) (22.94%)

Other audit programsw, $ millions . . . . . . . 1,777 1,620 1,518 1,992(29.64%) (24.98%) (16.44%) (19.53%)

Gross fiscal impact, $ billions . . . . . . . . . . . 6.00 6.50 9.20 10.20(100.00%) (100.00%) (100.00%) (100.00%)

Net fiscal impact, $ billions, after appeals(16%) and uncollectible (6%) . . . . . . . . . 4.68 5.00 7.00 7.96

(Table 1 is concluded on the next page.)


Note: In panel D, the percentage in parentheses represents the proportion of the dollar amountreceived from each source.

a Definitions not disclosed by CRA unless noted otherwise.b Canada Customs and Revenue Agency, CCRA Annual Report to Parliament 2002-2003 (Ottawa:

Canada Customs and Revenue Agency, 2003), 1-46, figures 5-1 and 5-2 (the annual reports areavailable online: http://www.cra-arc.gc.ca/agency/annual/menu-e.html).

c Canada Customs and Revenue Agency, CCRA Annual Report to Parliament 2003-2004 (Ottawa:Canada Customs and Revenue Agency, 2004), 1-30, figures 7 and 8.

d Annual Report 2001-2002, at 1-55, figures 9-1 and 9-2.e This includes criminal investigations of tax evasion and audits under the special enforcement

program of income obtained from illegal activity.f Canada Customs and Revenue Agency, CCRA Annual Report to Parliament 2000-2001 (Ottawa:

Canada Customs and Revenue Agency, 2001), 1-33.g Annual Report 2001-2002, at 1-40.h Annual Report 2002-2003, at 1-37.i Annual Report 2003-2004, at 1-22, exhibit 4.j na = not available.k Canada Customs and Revenue Agency, Comprehensive Discussion of Our Performance by Business

Line in Support of Accountability (Ottawa: Canada Customs and Revenue Agency, 2002), 14,exhibit 2 (accompanying the CCRA Annual Report to Parliament 2001-2002).

l Annual Report 2002-2003, at 1-43, figure 4-1.mAnnual Report 2003-2004, at 1-29, figure 6. The Annual Report 2002-2003 attributes the jump in

the bottom two series in this panel in that year to the refinement and inclusion of new criteriarather than to any real increase in non-compliance.

n Annual Report 2002-2003, at 1-44, figure 4-2.o Annual Report 2003-2004, at 1-24, exhibit 6.p Annual Report 2001-2002, at 1-43, figure 1-1. Information on taxable corporations is not

comparable to figures for later years due to a subsequent change in methodology.q Annual Report 2000-2001, at 1-27, exhibit 4.r Canada Customs and Revenue Agency, Annex: Supplementary Performance Information by Business

Line (Ottawa: Canada Customs and Revenue Agency, 2001), 2-44, exhibit 13 (accompanying theCCRA Annual Report to Parliament 2000-2001).

s Annual Report 2001-2002, at 1-53, figure 8-1.t Annual Report 2002-2003, at 1-47, figure 5-3.u Annual Report 2003-2004, at 1-31, figure 9.v This presumably includes individuals as well as corporations.w Other audit programs include tax avoidance, international tax programs, tax incentives, and

investigations.

TABLE 1 Concluded

Whether increasing audits and penalties is the best way to deal with non-compliance depends on the reasons why taxpayers comply or fail to comply. Broadlyspeaking, there are two classes of theories—economics-based theories, which em-phasize incentives, and psychology-based theories, which emphasize attitudes.6

Economic theories of compliance suggest that taxpayers “play the audit lottery”—

6 See Benno Torgler, “Speaking to Theorists and Searching for Facts: Tax Morale and TaxCompliance in Experiments” (2002), vol. 16, no. 5 Journal of Economic Surveys 657-83; James Almand Michael McKee, “Extending the Lessons of Laboratory Experiments on Tax Complianceto Managerial and Decision Economics” (1998) vol. 19, no. 4-5 Managerial and Decision Economics


that is, they make calculations of the economic consequences of different compliancealternatives (such as whether or not to evade tax) and the probabilities of detectionand the consequences of detection, and then they choose the alternative thatmaximizes their expected after-tax return (possibly after adjustment for the desiredlevel of risk). In contrast, the second class of theories assumes that psychologicalfactors—including moral and ethical concerns—are also important to taxpayers,and therefore taxpayers may comply even when the risk of audit is low. Of course,some taxpayer behaviour may be explained by the economic theories while somemay be explained by the psychological theories, and mixed explanations are alsopossible. A recent comment by a Canadian tax practitioner describes this variationwell with respect to the issue of transfer-pricing documentation:

The attitude of taxpayers varies considerably. Some comply to the fullest extent required,irrespective of whether the IRS is actively auditing them or their industry. Othertaxpayers essentially play the audit lottery and either do not prepare contemporaneousdocumentation or prepare the absolute minimum with the hope of avoiding theimposition of penalties.7

These theories are of more than academic interest since they imply differentconsequences regarding where the CRA should invest its resources. Economic theor-ies generally call for increased audits and penalties as the solution to complianceproblems. The policy prescriptions of psychological theories differ across theories,but these theories generally lead to policy prescriptions that de-emphasize auditsand penalties and instead focus on changing individuals’ attitudes toward the taxsystem. Such measures include increasing the system’s perceived fairness and mak-ing it easy to comply with the tax laws through superior Web site information,increased telephone assistance, and shorter lineups in CRA offices. Both theoriesgenerally support measures that result in reduced opportunity for non-compliance,such as third-party reporting.

The CRA’s interest in this topic is shown by its establishment of the ComplianceResearch and Strategic Analysis Division within the Compliance Programs Branch;this division has as one of its three mandates “comprehensive research and analysisof . . . factors that contribute to compliant and non-compliant behaviour.”8 Cur-rently, the division has a staff of 28, in contrast to perhaps 28,000 in the entire

259-75; Andrew D. Cuccia, “The Economics of Tax Compliance: What Do We Know andWhere Do We Go?” (1994) vol. 13 Journal of Accounting Literature 81-116; and Betty R. Jacksonand Valerie C. Milliron, “Tax Compliance Research: Findings, Problems, and Prospects” (1986)vol. 5 Journal of Accounting Literature 125-61, for reviews of these theories and associatedempirical studies.

7 R.D.M. Turner, “Transfer-Pricing Compliance Standards,” in Report of Proceedings of the Fifty-Fifth Tax Conference, 2003 Conference Report (Toronto: Canadian Tax Foundation, 2004),11:1-72, at 11:27.

8 Greg Maloney, “Compliance Research at the Canada Revenue Agency,” a presentation to theIRS Research Conference held in Washington, DC, June 2-3, 2004 (available on the Web athttp://www.irs.gov/taxstats/article/0,,id=122790,00.html).


CRA.9 However, the division has plans to “[i]dentify opportunities for partnering inresearch projects where there is mutual interest and benefit” (through joint projectswith other analytical areas within the CRA, which may potentially extend to othergovernment departments and academics) and to produce an annual report oncompliance.10

Alternative Research Approaches

Testing the economic- and psychology-based theories to see which can betterexplain taxpayer behaviour can be done in several ways. One approach is to studythe effects on compliance of actual changes in the laws, such as the introduction ofcivil penalties for tax practitioners. The problems with this approach are the lack ofsuch changes in adequate numbers and the statistical problem of sorting out thiseffect from the effect of simultaneous changes in other factors in the economy,such as economic growth and self-employment trends. Another approach, which isused in this study, is to use surveys and social-science experiments with actualtaxpayers with the goal of generalizing from the sample of taxpayers studied tooverall tax compliance behaviour for the whole country.11 Using this approach, thisstudy focuses on two research questions—(1) the best way to measure compliancebehaviour and (2) whether the theory of planned behavior (TPB) of Ajzen12 pro-vides a good explanation of tax compliance behaviour.

While measuring compliance behaviour appears to be a simple proposition, theproblem is to find a measure that allows the researcher to directly study howcompliance changes with a change in an underlying tax-related parameter, such asthe audit rate. Thus, directly asking participants about their past involvement in taxevasion is not sufficient. What is needed is a compliance indicator that can be repeat-edly measured as policy parameters change. One possibility is to ask hypothetical

9 There are 28,000 full-time equivalent staff members in the Tax Services Section of the formerCCRA: supra note 1, at 1-21 and 1-4.

10 Maloney, supra note 8, at 54.

11 Examples of studies that either focus on tax evasion in Canada or use Canadian participants intax evasion experiments are Janne Chung and Viswanath Umashanker Trivedi, “The Effect ofFriendly Persuasion and Gender on Tax Compliance Behavior” (2003) vol. 47, no. 2 Journal ofBusiness Ethics 133-45; Viswanath Umashanker Trivedi, Mohamed Shehata, and Bernadette Lynn,“Impact of Personal and Situational Factors on Taxpayer Compliance: An Experimental Analysis”(2003) vol. 47, no. 3 Journal of Business Ethics 175-97; Brian Erard, A Critical Review of theEmpirical Research on Canadian Tax Compliance, Working Paper 97-6 prepared for the TechnicalCommittee on Business (Ottawa: Department of Finance, September 1997); Neil Brooks andAnthony N. Doob, “Tax Evasion: Searching for a Theory of Compliant Behaviour,” in M.L.Friedland, ed., Securing Compliance: Seven Case Studies (Toronto: University of Toronto Press,1990), 120-64; and Roderick Hill and Muhammed Kabir, “Tax Rates, the Tax Mix, and theGrowth of the Underground Economy in Canada: What Can We Infer?” (1996) vol. 44, no. 6Canadian Tax Journal 1552-83.

12 Icek Ajzen, “The Theory of Planned Behavior” (1991) vol. 50, no. 2 Organizational Behaviorand Human Decision Processes 179-211.


questions about tax evasion behaviour (possibly with a case situation to describe thecontext in detail) and how it would differ if a specific change occurred. Alternatively,an experiment can be conducted in which the participants engage in tax-reportingsituations in a controlled setting and derive real monetary rewards based on theirdecisions. Within this experimental-economics framework, taxpayers receive a certainamount of income, choose how much to report to the tax authority, pay tax accord-ingly, are possibly audited, and, if they are found to have evaded tax, are required topay the tax evaded plus a penalty. The effects on compliance of changes in policyparameters can be measured by having the participants make repeated tax-reportingdecisions within the context of environments with different tax-related parameters.

Purposes of This Study

The choice between the hypothetical-situation and experimental-economics ap-proaches has often been made on a theoretical basis, with researchers choosing anapproach according to their assessment of the importance of real economic incen-tives in simulating real-life behaviour. This study posits instead that a more empiricalcriterion be applied—the indicator that is more strongly associated with past involve-ment in tax evasion. Our rationale is that since actual tax evasion is the ultimatevariable of interest, the proper indicator is the one that is most strongly correlatedwith it. By comparing the two indicators on this basis, this study provides someevidence on this issue.13

The second focus of this study is a determination of the precise chain of reasoningthat taxpayers use in making decisions about tax compliance. Although one candirectly ask people to explain their reasoning, tax-reporting behaviour is suffi-ciently complex to render this approach unworkable. Instead, the approach used isto postulate a specific psychological theory as to how individuals make decisionsand test whether it applies to tax compliance. This does not imply that the study

13 Studies by Henk Elffers, Henry S.J. Robben, and Dick J. Hessing, “On Measuring Tax Evasion”(1992) vol. 13, no. 4 Journal of Economic Psychology 545-67, and Dick J. Hessing, Henk Elffers, andRussell H. Weigel, “Exploring the Limits of Self-Reports and Reasoned Action: An Investigationof the Psychology of Tax Evasion” (1988) vol. 54, no. 3 Journal of Personality and Social Psychology405-13, have found that neither self-reports nor experimental measures of tax compliance arerelated to actual tax evasion behaviour. Both studies were of Dutch taxpayers and had the advantageof using tax evasion data from tax returns as their measure of actual tax evasion (as opposed to self-reported tax evasion, which we employed in our study). Hessing, Elffers, and Weigel compared thismeasure to taxpayers’ self-reports of tax evasion, while Elffers, Robben, and Hessing compared itto experimentally measured tax evasion as well. One possible problem with using tax-return-basedmeasures of tax evasion is that, as Elffers, Robben, and Hessing note, there was wide disagreementamong the tax officers on the tax evasion status of the individuals included in their study.

A second problem with the Elffers, Robben, and Hessing study is that the tax task in theirexperiment was embedded in a much wider business simulation context, yet their participants hadlittle real-life experience running similar businesses. Therefore, it is possible that the participantswere more intent on getting their business decisions “right,” and thus they may have failed to payadequate attention to the possibility of evading taxes. In any event, they may not have viewed theirtax compliance decisions in the experiment as reflective of their real-life tax compliance decisions.


excludes economic theories of tax compliance; instead, the statistical tests employ bothpsychological and economic variables. However, since it is not possible to consider allof the psychological theories for testing in one study, this study picks as a particulartheory for examination the TPB of Ajzen,14 which is a successor to the theory ofreasoned action (TRA) of Fishbein and Ajzen,15 Ajzen and Fishbein,16 and Ajzen.17

This theory is described below in the section on survey and experimental instruments.This study may be seen as an extension of previous TRA-TPB research to the

experimental-economics framework. Previous studies of the TRA and TPB in a taxcompliance context have used hypothetical situations18 and thus have not examinedwhether attitudes and intentions influence behaviour in the face of real monetaryconsiderations. On the other hand, previous experimental-economics work on taxhas typically excluded psychological considerations by refraining from using context-specific language in the instructions to the participants.19 In other words, thesestudies focused on people’s reactions to incentives without regard to the psycho-logical context of the incentives. Consequently, no attempt was made in these studiesto relate participants’ tax evasion decisions to their personality-related variables,

14 Supra note 12.15 Martin Fishbein and Icek Ajzen, Belief, Attitude, Intention and Behavior: An Introduction to Theory

and Research (Reading, MA: Addison-Wesley, 1975).16 Icek Ajzen and Martin Fishbein, Understanding Attitudes and Predicting Social Behavior (Englewood

Cliffs, NJ: Prentice-Hall, 1980).17 Icek Ajzen, Attitudes, Personality, and Behavior (Milton Keynes, UK: Open University Press,

1988), 112-45.18 Dennis M. Hanno and George R. Violette, “An Analysis of Moral and Social Influences on

Taxpayer Behavior” (1996) vol. 8, supplement Behavioral Research in Accounting 57-75; CindyBlanthorne, “The Role of Opportunity and Beliefs on Tax Evasion: A Structural EquationAnalysis” (PhD dissertation, Arizona State University, 2000); and Donna D. Bobek andRichard C. Hatfield, “An Investigation of the Theory of Planned Behavior and the Role ofMoral Obligation in Tax Compliance” (2003) vol. 15 Behavioral Research in Accounting 13-38.

19 Typically, neutral terms such as “fee” for income, “check” for audits, and “shortfall payment”for penalties are used in these studies. Bloomfield and Libby’s research in the area of financialreporting shows that experimental markets results are different if one uses familiar terms thathave meaning rather than artificial terms that are unfamiliar to the participants: Robert R.Bloomfield and Robert R. Libby, “Market Reactions to Differentially Available Information inthe Laboratory” (1996) vol. 34, no. 2 Journal of Accounting Research 183-207. Tax evasion decisionsin particular have strong ethical and moral implications, which are typically absent in accounting-related tasks. These implications can also influence participants’ tax compliance behaviour in thelaboratory. Therefore, the issue of using contextual (actual tax-related) terminology versus usingneutral terms is of special significance in the tax evasion context as compared with other contexts.

Previous studies that have used tax terminology in an experimental-economics frameworkinclude Julie H. Collins and David R. Plumlee, “The Taxpayer’s Labor and Reporting Decision:The Effect of Audit Schemes” (1991) vol. 66, no. 3 The Accounting Review 559-76, and DonaldV. Moser, John H. Evans III, and Chung K. Kim, “The Effects of Horizontal and ExchangeInequity on Tax Reporting Decisions” (1995) vol. 70, no. 4 The Accounting Review 619-34,among others. However, none of these studies attempted to link individuals’ personality-relatedvariables, such as attitudes and intentions, to their tax compliance decisions in the laboratory orattempted to relate the latter to the participants’ actual compliance decisions.


attitudes, and intentions.20 To allow for the incorporation of this set of potentialdeterminants of tax evasion, we made use of tax-specific language by urging ourparticipants, via repeated verbal instructions, to act in a manner similar to the wayin which they would act if placed in a similar taxpaying situation in real life.

In the next section, we discuss our work on the measurement of tax compliancebehaviour. We describe our sample, our survey and experiment instruments, andour results. We then report on our test of one psychological theory of tax compli-ance, the TPB; we discuss the portions of our survey that relate to this theory, andthen our results. Finally, we conclude with the contributions and limitations of thisarticle, together with suggestions for future research.

M E A S U R I N G T A X C O M P L I A N C E B E H AV I O U R

Overview of the Survey and Experiment

Hill and Kabir21 report that participation in the underground economy (and, byextension, tax evasion in Canada) is especially high among low-income individuals(including students, the unemployed, and those receiving social assistance), youths,and unmarried persons. In this context, Jackson and Milliron22 observe that taxevasion studies should focus on groups of interest rather than a diverse set ofparticipants. Consequently, our study focuses on students recruited from a mid-size Canadian university, McMaster University in Hamilton.

We held a three-hour session to obtain demographic data, administer the surveyinstruments, and conduct the experiment.23 Our survey and experiment instru-ments (forms for the participants to fill out) covered tax compliance behaviour and

20 Outside the TRA-TPB context, the study by Julie H. Collins, Valerie C. Milliron, and DanielR. Toy, “Determinants of Tax Compliance: A Contingency Approach” (1992) vol. 14, no. 2 TheJournal of the American Taxation Association 1-29, is a good example of the use of personality-related variables to explain hypothetical tax evasion decisions.

21 Supra note 11.

22 Supra note 6.23 The sequence of administering the different tasks was as follows: demographic data were

elicited first, the survey was completed next, the experiment was administered third, and thedefining issues test (DIT), which is discussed below, was completed last. Refreshments wereprovided to the participants during the course of the three hours to maintain energy levels andto reduce fatigue. The survey instrument included questions to elicit participants’ attitudesand intentions as well as hypothetical compliance decisions. Therefore, the order of elicitinginformation may have influenced participants’ subsequent behaviour in the experiment.However, we are not aware of any a priori reason why such influence should differ betweenhypothetical and experimentally measured compliance decisions. Given that our main focus isthe comparison between these two types of compliance decisions, we believe that the influenceof order, if any, does not pose a major threat to our findings. Further, the DIT measures innatemoral reasoning, which, theoretically, is not susceptible to change over short periods of time( James R. Rest, Robert Barnett, Muriel Bebeau, Deborah Deemer, Irene Getz, Yong Lin Moon,James Spickelmier, Stephen J. Thoma, and Joseph Voloker, Moral Development: Advances inResearch and Theory (New York: Praeger, 1986), 28-58). Therefore, we do not apprehend aproblem as a result of administering the DIT last.


psychological variables suggested by the TPB.24 The instruments are available fromthe McMaster Experimental Economics Laboratory Web site.25

Participants’ compensation for completing the above tasks included a show-upfee of $10, earnings from the experiment, and $20 to complete the last surveyinstrument, the defining issues test (DIT, discussed below). Parameters for thelaboratory sessions were structured in such a way that, on average, participantswould earn approximately $15 per hour. For the 10 decision rounds, 41 partici-pants were provided with an income of $1 (5,000 Lira) per round and 38 participantswere provided with an income of $1.80 (9,000 Lira) per round. The special “Lira”currency, in which 5,000 Lira equal $1 Canadian, was used to ease the calculationof taxes and penalties to be paid by the participants by allowing for easy rounding offigures. If Canadian dollars had been used, participants may have had to deal withdecimal amounts. The use of laboratory currency, while not required, is commonin experimental-economics studies.

To maintain participants’ anonymity and to reduce any implicit pressure partici-pants may have felt to provide the “right” answers, participants were identified onlyby unique identification numbers. Participants were further assured of the confi-dentiality of the study through a cover letter, which promised that only the principalresearchers would have access to their data.

Consistent with the fact that all of our participants were students, the mean(that is, average) age of our sample was 24.4 years, with a range of 19 to 54 years.Of our 79 participants, 65 were undergraduates and 14 were MBA students. Oursample was almost evenly matched in terms of gender, with 40 females and 39 males.Fully 96 percent of our participants had tax-filing experience. One individual pro-vided incomplete information and was therefore dropped from the sample, resultingin a usable sample of 78 participants.

Compliance Measures

Two measures of past evasion were elicited from participants. The first question weasked was, “Did you, when filing last year’s tax return, underreport your income orreport unwarranted deductions?” The second question we asked was, “Have youever underreported income or reported unwarranted deductions on any previousyear’s tax return?” The answers allowed were “yes” and “no.” (In statistical work abinary scale is used, with a “yes” response coded as 1 and a “no” response coded as 0.)

24 The instruments used to elicit participants’ attitudes, social norms, perceived behaviouralcontrol, and intentions are based on different instruments that were pre-tested and used byHanno and Violette, supra note 18, and Blanthorne, ibid. However, we did not use the actualquestions used by Hanno and Violette, since the authors of that study could not provide uswith a copy of their instrument. Instead, we came up with questions similar to those used byHanno and Violette based on the descriptions provided about each question in their paper.

25 See http://socserv.socsci.mcmaster.ca/econ/mceel/papers/taxcomp-inst.pdf and http://socserv.socsci.mcmaster.ca/econ/mceel/papers/taxcomp-surv.pdf. The “Defining Issues Test,”which is discussed below, is not included because it is copyright material.


We also asked several questions about hypothetical compliance. The first wasworded as follows:

Assume that you have earned an income of $2,500. The tax rate is 30 percent. There isa 0 (Zero) percent chance that your tax return will be audited and the full extent of yourunreported income will be detected by Revenue Canada. Please indicate below howmuch of the amount of $2,500 earned by you that you will report to Revenue Canada.

The second question was similar, but it allowed for audits:

Assume that you have earned an income of $2,500. The tax rate is 30 percent. Thereis a 25 percent chance that your tax return will be audited and the full extent of yourunreported income will be detected by Revenue Canada. On detection, not only willyou have to pay the taxes owed on the unreported tax, but you will also have to pay apenalty equal to 1.5 times such unpaid taxes. Please indicate below how much of theamount of $2,500 earned by you that you will report to Revenue Canada.

The compliance measures corresponding to these questions—hypothetical-situation compliance at 0 percent audit and hypothetical-situation compliance at25 percent audit—were the proportions of the $2,500 in income reported to theCRA (“Revenue Canada” at the time of the experiment) in the two situations.

Participants also responded to four case situations of four to seven sentenceseach, designed to elicit the likelihood that the participant would not report to theCRA each of four different types of income: restaurant tips and three types of part-time self-employment income (artist, computer consultant, and home handyman).26

Participants’ responses regarding this likelihood were obtained on a Likert-typescale ranging from -3 to +3, with -3 labelled “unlikely” and +3 labelled “likely,”and 0 being the midpoint. Responses to all four case scenarios were highly corre-lated to each other;27 thus, the four variables were combined into one variable,

26 It is not clear what audit rate the survey participants had in mind when answering this question.As noted supra note 5, an audit rate for individuals is a murky concept. Even if the averageaudit rate for the country of about 1.3 percent (shown in panel A of table 1) is accurate, the auditrate is not uniform over the entire population. The CRA targets those posing a higher risk ofevading taxes, and our survey participants appear to be in a higher-risk category. On the otherhand, since our participants are all college students and are by and large young, they may face alower audit rate because the tax revenues recoverable may be meagre. Also, it is quite possiblethat most people think that a CRA tax audit is a random event, although they have no idea of theprobability of this random event. Further, even if they know the true audit rate, they may act asif it is larger that it really is: James Alm, Gary H. McClelland, and William D. Schulze, “WhyDo People Pay Taxes?” (1992) vol. 48, no. 1 Journal of Public Economics 21-38.

27 The associated Cronbach’s alpha, which is a measure of internal consistency based on the averageinter-item correlation, is a very high 0.814. The four cases, though in general patterned on casesused in prior studies, were specially prepared by us for our research project and have thereforenot been pre-tested. These cases are short and may therefore at first sight appear spare and notrich enough to capture taxpayers’ mental taxpaying models. However, our statistical results


termed “responses to cases,” which was used in further data analysis.28 Response tocases may also be considered an indicator of hypothetical compliance, resulting in atotal of three measures of this concept.

We measured compliance in the face of monetary incentives using experimental-economics techniques. The instructions given to each participant explained that heor she had received some income and had to decide what amount to report to RevenueCanada. We asked our participants to behave identically to what their compliancebehaviour would be if placed in real-world tax compliance contexts with similareconomic consequences.29 The penalty for understating income was stated to be150 percent of the tax evaded. There were 10 repetitions (“rounds”) for each partici-pant, with the only difference across rounds being the audit rate. This rate was setat 0 percent for the first five rounds and 25 percent for the next five rounds.30 Theaverage proportion of income reported over the five rounds under each treatmentwas used as the indicator of experimentally measured compliance for that auditrate. In the tables below, the two variables are called “experimentally measuredcompliance at the 0 percent audit level” and “experimentally measured complianceat the 25 percent audit level,” respectively.31

In the five rounds with the audit rate of 25 percent, all participants knew beforemaking their income-reporting decisions for those rounds that there was a one infour (or 25 percent) chance that they would be audited in a particular round and thattheir chance of getting audited in each round was random and independent of whathappened in a previous round. Four numbered balls in a can were used to deter-mine whether or not participants were audited in each of those rounds. One of theparticipants was asked at the beginning of the five rounds to identify one of the ballsas the audit ball by randomly picking one of them. Thereafter, in each of the fiverounds, after all participants had submitted their income tax reports, one of the

indicate otherwise. While the participants’ responses to all four case scenarios load onto onefactor, there are significant differences in the average responses, which are noted below in the“Results” section.

28 Factor analysis suggested that all four variables loaded onto a single component (results notpresented); their factor score is the variable described above as “responses to cases”. Wheneverrequired, we used factor scores from the related factor analysis, and not simple averages of theunderlying indicators, during further data analysis.

29 To make such a requirement salient to the participants, they were instructed verbally instead ofvia written instructions to act as they would in a real tax setting. Further, these instructions wererepeated throughout the experiment to ensure that the participants did not forget this point.

30 We started with the 0 percent audit rate because we wanted to first elicit participants’ behaviourin the absence of audits, when they had no prior experience of being audited in the experiment.To mitigate any ordering effect, we asked the participants to provide their hypotheticalcompliance decisions assuming a 25 percent audit rate first and a 0 percent audit rate next.Nonetheless, given the large difference in the two rates—0 to 25 percent—we do not believethat the order of administering the two levels had a qualitative effect on our results.

31 Similar results were obtained when data from the last round in each set—that is, the 5th and10th rounds respectively—were used instead in our statistical analysis.


participants was asked to pull out one of the four balls at random and withoutlooking. If the ball picked up was the designated audit ball, then all participantswere audited in that round. Otherwise, no participant was audited in that round.Thus, the audit procedure was conducted in full view of the participants so as tomake the risk of being audited tangible and real to them.

For the experimental-economics work, some individuals had a higher income perround than others. Participants provided with 5,000 Lira per round had to pay taxesat 10 percent, while participants at the 9,000-Lira level had to pay taxes at 50 percent.Therefore, at full compliance the after-tax income for both levels of income wouldbe the same—4,500 Lira.32 As a consequence, however, a limitation of our study isthat we cannot distinguish statistically between differences caused by changes inincome and differences caused by changes in tax rates.

Results

Levels of ComplianceTable 2 shows the levels of compliance revealed by participants. In terms of actualcompliance, 13.9 percent of all participants indicated that they did not report all oftheir income in the previous year, and 17.9 percent indicated that they did not reportall of their income in any past year. For the hypothetical situations, an average of57 percent of income was reported when there was no chance of an audit, and76 percent was reported when there was a 25 percent chance of an audit.33 In theexperimental-economics decision rounds, where money was at stake, only 55 per-cent of income was reported when there was no chance of an audit; 62 percent wasreported when there was a 25 percent chance of an audit.34 As previously noted,tax-specific language was used in these controlled environments and participantswere told via verbal instructions to act as they would in a real tax setting.

It is perhaps a concern that the average levels of actual self-reported complianceon tax returns are greater than the average levels of both hypothetical-situationcompliance and experimentally measured compliance. A potential explanation forthe difference may be that third-party reporting occurs in real life but was notimplemented in this experiment, i.e., audit rates are much greater than 25 percentin real life for employment and other types of income. Alternatively, the participantsmay have underreported their actual degree of tax evasion. Perhaps compliance

32 Different levels of income with different tax rates were provided to these two groups todetermine whether compliance would vary between the two groups because of the difference intax rates despite the fact that both groups would be equally well-off when they were totallycompliant.

33 Of our participants, 65.8 percent and 39.2 percent did not fully report their hypotheticalincome when the audit rates were zero and 25 percent, respectively.

34 Of our participants, 77.2 percent and 74.7 percent did not fully report their income when theaudit rates were zero and 25 percent, respectively, in the experiment.


behaviour in the laboratory reflects compliance in real-life situations where there islittle chance of being caught, as in the case of income from tips or casual labour.

The reported levels of hypothetical-situation and experimentally measured com-pliance shed some light on the predictive validity of economic theories of compliance.Since the penalty for understating income was stated to be 150 percent of the taxevaded, the participants were essentially choosing between a payment of the truetax with certainty and a gamble of no tax if they were not audited and 250 percentof the tax if they were audited. This implies that a 40 percent audit rate would havebeen required to make a participant who was maximizing expected income to beindifferent between compliance and non-compliance. With any lower audit rate, theparticipant would not have complied and hence would have paid no tax. Since boththe hypothetical-compliance measures and the experimental-economics measureshad a maximum audit rate of 25 percent, a participant who was maximizing expectedincome would have had 100 percent non-compliance. Thus, compliance was muchhigher than would be predicted by a simple economic theory of decision making.Presumably, psychological factors such as ethics must explain the higher levels ofcompliance that were observed.

TABLE 2 Descriptive Statistics on Attitude, Intentions, PersonalityVariables, and Different Measures of Compliance

StandardVariable n Min. Max. Mean deviation

ComplianceHypothetical-situation compliance at 0% audit . . . . . . 79 0.000 1.000 0.566 0.406Hypothetical-situation compliance at 25% audit . . . . . 78 0.000 1.000 0.763 0.342Experimentally measured compliance at 0% audit . . . . 79 0.000 1.000 0.551 0.373Experimentally measured compliance at 25% audit . . . 79 0.000 1.000 0.617 0.366Response to cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 -2.603 1.380 0.000 1.000Non-compliance in previous year . . . . . . . . . . . . . . . . . 79 0.000 1.000 0.139 0.348Non-compliance in prior years . . . . . . . . . . . . . . . . . . . 78 0.000 1.000 0.179 0.386

AttitudesSubjective norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 -2.129 2.158 0.000 1.000Duties of a citizen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 -1.932 3.223 0.000 1.000Monetary considerations . . . . . . . . . . . . . . . . . . . . . . . . 78 -2.165 2.686 0.000 1.000Advantage of the system . . . . . . . . . . . . . . . . . . . . . . . . . 78 -1.828 3.453 0.000 1.000Penalties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 -2.525 3.937 0.000 1.000Third-party reporting . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 -3.104 3.133 0.000 1.000

IntentionsIntent to comply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 -2.852 1.267 0.000 1.000Intent to overstate deductions . . . . . . . . . . . . . . . . . . . . 79 -3.000 3.000 -0.392 1.996

Personality variablesA score . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 0.000 11.000 3.096 2.811P score . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 10.000 66.700 34.980 13.682

Note: See the text for the development of these variables.


Correlations Among Compliance MeasuresAs noted in the introduction to this article, researchers need an indicator of taxcompliance that can be varied within the survey or experiment for different valuesof the policy parameters. This allows the researcher to determine, for example, theeffect of changes in audit rates on compliance behaviour. For the indicator to havevalidity in predicting real-life behaviour, it should be highly correlated with pastcompliance behaviour, as measured in our study by two variables: non-compliancein the previous year and non-compliance in any past year.

This study included two potential indicators: hypothetical tax compliance, meas-ured by three variables, and experimentally measured tax compliance, measured bytwo variables. Table 3 shows that the three variables for hypothetical compliance allhad low correlations with actual tax return non-compliance, with just one havingany reasonable degree of statistical significance: response to cases had a 0.22 correla-tion with actual non-compliance in the previous year. Even this correlation lacksmuch practical significance, because it implies that response to cases explained lessthan 5 percent of the variation in actual non-compliance in the previous year.35 Allof the correlations had the expected sign: positive for response to cases and negativefor hypothetical-situation compliance at the 0 percent and 25 percent audit rates.

In contrast, table 3 shows that experimentally measured tax compliance was muchbetter correlated with actual compliance—all correlations were the right sign (thatis, increased experimentally measured compliance was associated with increasedactual compliance) and all were statistically significant at various levels, and thepercentage of variance explained varied from 4 percent to 13 percent. The strongestcorrelations were with actual tax-return non-compliance in the previous year; thisis the most important measure of real-life behaviour because it reflects the mostrecent actual decision-making.

Still, there is obviously much variance left to be explained. Part of the problemmay be that the measure of actual non-compliance is a variable with values of 0 or1, which cannot be expected to explain a high percentage of the variance of acontinuous variable.36 Another issue, as stated above, is that responses on actual non-compliance in the past may reflect non-compliance both in the presence and absenceof third-party reporting, while the different measures of non-compliance elicited inthis study may reflect non-compliance in the absence of any third-party reporting.

Note that the experimental-economics compliance measures used in this studywere elicited using tax language and repeated reminders to our participants tobehave as they would in real life (as discussed in the introduction to this article).Thus, the converse finding of Elffers et al.37 that their experimentally determined

35 The square of the correlation is the proportion of variance explained. This has a minimum of0 percent and a maximum of 100 percent.

36 Future researchers might ask participants what percentages of last year’s income and income inprevious years they reported to the tax authority.

37 Elffers, Robben, and Hessing, supra note 13.


tax compliance was unrelated to actual tax evasion behaviour may perhaps beexplained by the fact that that study used a general business context and did nothighlight its link to tax compliance. The previous discussion suggests that the useof tax terminology and instructions to act in a manner consistent with real life mayproduce tax compliance measures that are more closely connected to real-worldbehaviour.

Compliance by SituationParticipants’ average responses to the case employment situations suggested thatthe ranking of the employment situations in which they were most to least likely toevade tax on the income was as follows: restaurant worker (mean 1.34), homehandyman (mean 1.24), artist (mean 0.71), and computer consultant (mean 0.43).Thus, it appears that the participants were more sympathetic to and more inclinedto behave like the characters in the cases when there appeared to be a genuinereason for such tax evasion (for example, when the evader was a poor student fundingher university education through restaurant tips or a home handyman supplement-ing his income to support his wife and four children).38

TABLE 3 Parametric Correlations Between DifferentMeasures of Compliance (N = 78)

Non-compliance in Non-compliancethe previous year in any past year

Correlations relating to hypothetical-situationcompliance measures

Hypothetical-situation compliance at 0% audit . . . . . -0.057 -0.089(0.3)a (0.8)

Hypothetical-situation compliance at 25% audit . . . . 0.022 -0.045(0.0) (0.2)

Response to cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.218b 0.123(4.8) (1.5)

Correlations relating to experimentally measuredcompliance measures

Experimentally measured compliance at 0% audit . . . -0.368c -0.221b

(13.5) (4.9)Experimentally measured compliance at 25% audit . . . -0.301c -0.226d

(9.1) (5.1)

Note: See the text for the development of these variables.a Figures within brackets represent percentage of variance explained.b Significance at the 10% level.c Significance at the 1% level.d Significance at the 5% level.

38 Some of these differences are statistically significant. When paired-sample t-tests were conducted,the average response relating to the restaurant worker case was significantly different (p < .001)from the average response of the participants relating to either the computer consultant or


T E ST I N G A T H E O RY O F T A X C O M P L I A N C E

Variables Suggested by the Theory of Planned Behaviour

Ajzen’s theory39 models behaviour as depending on behavioural intention, andbehavioural intention in turn as depending on three factors: (1) attitude towardthe behaviour, (2) subjective norms, and (3) perceived behavioural control. Thistheory is diagrammed in figure 1 (but note that the bottom two arrows relating toethics are explained below). Attitudes relate to one’s own personal views about thebehaviour. Subjective norms are a function of an individual’s perception that one ormore relevant people would approve of a particular behaviour, and the extent towhich the individual’s behaviour will be influenced by that. Perceived behaviouralcontrol refers to an individual’s perception regarding the extent of control that heor she has over a particular behaviour. These concepts will become clearer as wediscuss below the questions that are used to measure them.

Perceived behavioural control is modelled as capable of influencing behaviourdirectly as well as indirectly through its effect on intentions. The indirect linkrepresents the motivational influence of control on behaviour. To the extent thatpeople believe that they have limited control over performing a particular behav-iour, their intention to perform the behaviour may be decreased. The direct linkfrom perceived behavioural control to behaviour represents the actual control thatan individual has over behavioural performance.

Participants were asked two questions concerning each of 23 aspects of attitudes,social norms, and perceived behavioural control that relate to tax compliance.They first were asked how much they valued each of the beliefs, and then they wereasked to rate the degree to which (or the probability with which) the associatedbelief was likely to occur or ensue.40 Two sample questions related to social norms:

■ “How much do you care whether your current or future employer approvesor disapproves of your correct reporting of your income and deductions onyour tax return? [with answer endpoints “not at all” and “very much”].”

■ “If I complied with the federal and provincial income taxes (if I correctlyreported my income and deductions on my tax return), my current or futureemployer would: [approve, disapprove].”

Similarly, a pair of questions related to attitudes:

■ “Fulfillment of my personal, moral, and ethical obligations as a consequenceof my complying with the federal and provincial income tax laws (correctly

artist cases; the average response to the computer consultant case was significantly differentfrom the average response to the home handyman case (p < .001); and the average response tothe artist case was significantly different from the response to the home handyman case.

39 Supra note 12.

40 This is called “expectancy” in the literature.


reporting my income and deductions on my tax return) is: [with answerendpoints “bad” and “good”].”41

■ “Complying with the federal and provincial income tax laws (correctly re-porting my income and deductions on my tax return) will result in thefulfillment of my personal, moral, and ethical obligations: [unlikely, likely].”

Responses to both types of questions were measured on a 7-point Likert scalefrom -3 to +3. Zero (0), the midpoint in the range, indicated a neutral response;negative values indicated a negative response (such as “unlikely” or “not at all”);and positive values indicated a positive response (such as “likely” or “very much”).

For each pair of questions, an indicator was constructed by multiplying thenumbers representing the answers to the two questions.42 Since many of the 23indicators were similar to each other, and including them all in our statisticalanalysis was not feasible given our sample size, we used factor analysis to reducethese 23 indicators into 6 components.43 Of the 23 indicators, 20 load onto (that is,

FIGURE 1 Theory of Planned Behaviour with Moral Reasoning

IntentionPerceived

behaviouralcontrol

Subjectivenorms

Attitude

Ethics

Behaviour

41 For this question, a clearer indication of how much the participant valued the belief might havebeen obtained by using answer endpoint words such as “important” and “unimportant” insteadof “good” and “bad.” Nonetheless, the questions used by Hanno and Violette, supra note 18,were employed because they had been pretested in that previous study. It appears that ourparticipants understood the questions we used and answered them correctly, since participants’responses to these questions were in the expected direction and the indicators that are similarloaded significantly onto similar factors.

42 Before multiplying, one of the two Likert scales in the pair, the likelihood scale, was transformedfrom the range [-3, 3] to the range [1, 7]. This transformation avoids inaccurate indicator scores,since both (-3) times (-3) and 3 times 3 equal 9. After the transformation, the indicator for the(-3, -3) pair becomes -21 and the indicator for the (3, 3) pair becomes +21.

43 We consistently adopted a cutoff of one for eigenvalues while identifying factors during factoranalysis. Exploratory factor analysis was necessary to identify which of the 23 indicators needed


are strongly correlated with) 6 principal components. The 6 components have beenlabelled as follows.44 For convenience, the factor from figure 1 to which each ofthese components relates is listed first before the components are listed.45

Subjective Norms Factor■ “Subjective norms” component. Six indicators load onto this component: the

questions relating to the importance to tax compliance decisions of the ap-proval of such decisions by one’s friends, tax preparer, peers, spouse, employer,and family.

Attitudes Factor■ “Duties of a citizen” component. Five indicators load onto this component: the

importance to the tax compliance decisions of fulfilling one’s moral andethical obligations, funding the government, fulfilling the duties of a citizen,recognition of one’s good ethics by the community, and the presence ofrecords (for example, computer records or receipts).

■ “Advantage of the system” component. Four indicators load onto this compon-ent: the importance to tax compliance decisions of one’s perceptions ofhaving taken advantage of the system or of factors necessary for doing so—specifically, being able to brag about one’s actions, knowledge of the incometax system, a feeling of beating the system, and the effort required to prepareone’s tax return.

■ “Monetary considerations” component. Two indicators load onto this compon-ent: the prospect of receiving a refund and the prospect of having extra cashconsequent to one’s actions.

to be retained. Further, reliability analysis using Cronbach’s alpha indicated that this groupingof the 20 indicators across the six components was reasonable (see infra note 44). We usedexploratory factor analysis to reconfirm the validity of the instruments used in our study.Specifically, we believe that exploratory and not confirmatory factor analysis (structuralequations modelling) was appropriate in our study given that (1) our study was the first to useCanadian participants, whereas earlier research that used these indicators was conducted usingUS participants; (2) this was the first use of the questionnaire we devised (based on thedescription provided by Hanno and Violette, supra note 18), and (3) Barbara G. Tabachnickand Linda S. Fidell, Using Multivariate Statistics, 3d ed. (New York: HarperCollins College,1996), 715, note that a sample size of at least 200 is required for small to medium-sized modelsin order to use structural equations modelling, which is a confirmatory technique. Whilesample size is also an issue for exploratory factor analysis, the requirements are not as stringent.

44 The Cronbach’s alpha values relating to the first five components are respectively 0.863, 0.849,0.659, 0.763, and 0.374.

45 For each of the six factors above, the related factor scores, which are normalized with a meanof 0 and calculated using the regression method, were used in further data analysis. Theindividual data that we analyzed were therefore deviations from the mean of the factor scorefor each factor.


Perceived Behavioural Control Factor■ “Penalties” component. Two indicators load onto this component and relate to

the importance to tax compliance decisions of legal or social penalties ensu-ing from one’s actions: the possibility of penalties (such as fines or jail time)and making friends feel awkward as a result of one’s actions.

■ “Third-party reporting” component. A single indicator loads onto this component.

Table 2 provides descriptive statistics on these six components.

Tax Compliance Intentions

Following Hanno and Violette,46 we elicited five measures of intention to complyor to misreport deductions. Intentions are distinguished from behaviour in that theformer are responses to questions that begin, “I will try to” or “I intend to” and arenot specifically about past behaviour. All five were measured using a Likert-typescale ranging from -3 to +3, with -3 labelled “unlikely” and +3 labelled “likely”that the participants would report their income or overstate their deductions. Thewording of four of these measures is very similar, and in factor analysis they load ontoa single component (the associated Cronbach’s alpha is 0.791); this component wasnamed “intent to comply” and was used in further statistical tests. The last meas-ure, being unique, was retained in its original unmodified and unnormalized formand was labelled “intent to overstate deductions.” In summary, therefore, we usetwo measures of intention—intent to comply and intent to overstate deductions.

Ethics

Ajzen47 notes that the TPB is open to the inclusion of additional predictor variablesif they improve the prediction of intention or behaviour after the original variableshave been taken into account. In the tax compliance context, ethics appears to be anatural choice as an additional predictor that influences behaviour both directly andthrough behavioural intentions. Measures of this additional construct are discussedbelow. The bottom two arrows in figure 1 illustrate this concept.

One dimension of ethics is moral reasoning. We used the P score of the definingissues test (DIT) to investigate the impact of moral reasoning on tax compliance.The DIT is grounded in the theory of the stages of moral reasoning first proposedby Piaget48 and later formalized by Kohlberg.49 Rest50 developed the DIT to assess

46 Supra note 18.47 Supra note 12.48 Jean Piaget, The Moral Judgment of the Child, trans. Marjorie Gabain (New York: Free Press,

1965) (originally published in 1932).49 Lawrence Kohlberg, “Stage and Sequence: The Cognitive-Development Approach to

Socialization,” in David A. Goslin, ed., Handbook of Socialization Theory and Research (Chicago:Rand McNally, 1969), 347-480.

50 James R. Rest, Development in Judging Moral Issues (Minneapolis: University of MinnesotaPress, 1979).


participants’ moral reasoning in terms of the Kohlberg model, and its reliabilityhas been tested in hundreds of research studies.51 The P score represents the extentof the post-conventional level of moral development attained by each individual.52

The predicted relationship with compliance is positive, so that participants withhigher P scores may be expected to show higher levels of compliance.

Studies of tax compliance in the laboratory have also used another dimension ofethics: whether the participant exhibits an “anti-establishment orientation, a pointof view which condemns location and existing social order for its arbitrariness orits corruption by the rich for exploitation of the poor. . . . [It is] possibly a transitionphase between conventional morality and principled morality.”53 This is measuredby the A score on the DIT. If we assume that the government and its tax authoritiesrepresent the “establishment,” then intuitively the A score may have some relation-ship to tax compliance. The expected relationship is negative, so that participantswith higher A scores should have lower levels of compliance. Both the P and A

51 Rest et al., supra note 23, at 28-58.

52 Ibid., at 185-200. Kohlberg’s stage theory states that individuals make moral judgments based ona concept of justice that develops as they mature. Kohlberg, supra note 49, at 382-91, identifiesthree levels of moral development (pre-conventional, conventional, and post-conventional),with each level including two distinct stages of moral reasoning. In the pre-conventional level,moral decisions are motivated by fear of punishment (stage 1) or from self-interest (stage 2). Atthe conventional level, moral decisions are influenced by the social group, so that the individualacts to please or help others (stage 3), or to fulfill societal, legal, or religious norms—that is,rule-based reasoning (stage 4). Finally, at the post-conventional level, individuals make ethicalchoices based on concepts of justice, such as individual rights and socially accepted standards—that is, a social contract (stage 5)—or from universal ethical principles as defined by thedecision maker’s conscience (stage 6). Thus, the moral reasoning stages involve one’sperception of the justice of the “self ” versus “others” relationship.

The DIT consists of either three or six cases relating to ethical dilemmas. The six-caseextended DIT was used in this study. One example of these cases concerns a husband who isfaced with an ethical dilemma stemming from his wife’s life-threatening disease. The husbandhas to decide whether or not to break into the pharmacy of a price-gouging pharmacist to steala very rare drug invented by that pharmacist; this drug can cure his wife of her life-threateningdisease. After reading each case, participants were required to respond to many questionsrelating to that case, including whether or not the action contemplated in the case should becarried out. Participants’ responses were captured on a machine-readable scoring sheet. Thescoring sheets with the participants’ responses were sent to the Center for the Study of EthicalDevelopment at the University of Minnesota for scoring. The P score and the A score for eachparticipant was generated by the Center and sent back to us. While an accounting-specific DITexists (see Linda Thorne, “The Development of Two Measures To Assess Accountants’Prescriptive and Deliberative Moral Reasoning” (2000) vol. 12 Behavioral Research in Accounting139-69), a tax-specific DIT does not, thereby motivating us to use the generic DIT of Rest,supra note 50.

53 James R. Rest, Guide for the Defining Issues Test (Minneapolis: Center for the Study of EthicalDevelopment, 1987), 4.2.


scores from the DIT have been used in prior research to explain tax compliancebehaviour in the laboratory.54

Results

Recapping the theory presented above and pictured in figure 1, we had two propo-sitions to test:

■ Hypothesis 1: Tax compliance behaviour is determined by three factors: taxcompliance intention, perceived behavioural control, and ethics. Attitudesand subjective norms should not directly affect tax compliance behaviour.

■ Hypothesis 2: Tax compliance intention is determined by four factors: attitudes,subjective norms, perceived behavioural control, and ethics.

Each of these hypotheses was tested by statistically estimating a regression equa-tion and checking whether the variables relating to the determining factors listedabove were statistically significant at the 5 percent level.55

54 Steven E. Kaplan, Kay J. Newberry, and Philip M. Reckers, “The Effect of Moral Reasoningand Educational Communication on Tax Evasion Intentions” (1997) vol. 19, no. 2 The Journalof the American Taxation Association 38-54; and Trivedi, Shehata, and Lynn, supra note 11.

55 Given that our sample size was only 78 and the number of potential variables and interactionswas large, the following procedures were adopted to estimate the regression equations. First,stepwise regressions were estimated without interactions. Cutoffs of 0.20 and 0.25 were usedfor including and excluding variables, respectively. The variables included initially in a modeldepended on the nature of the associated dependent variables. In the models with hypotheticalmeasures of compliance as the dependent variables, the independent variables included initiallyin the model were all six principal components, the two measures of intent, the variable toidentify whether the participant was an undergraduate or an MBA, the P score, and the A score.When the dependent variables were the experimentally measured compliance measures, thesevariables were supplemented by a variable to identify the income provided—5,000 or 9,000 Lira.In the models with two data points per participant, the audit rate was included as an additionalindependent variable. Similarly, in the models in which the dependent variables were the intentmeasures, the independent variables initially included were the six principal components, thevariable to identify whether the participant was an undergraduate or an MBA, the P score, andthe A score. Second, using only the independent variables found to be significant for eachmodel in the above procedure, we added interaction terms so as to include not only the maineffects but also all possible higher-order interactions between such main effects. Finally, allinteractions insignificant at or above the 0.25 level were eliminated, with their sums of squaresbeing pooled with the error term. Only these final equations are reported in the tables.

Eliminating insignificant variables using stepwise regressions is also appropriate given that(1) our study is not a test of any one particular theory, and (2) variables included in thesestepwise regressions are motivated by more than one theory. Further, the fact that only someand not all variables motivated by any one theory are significant attests to the appropriatenessof using stepwise regressions initially to identify significant variables. Finally, the use ofstepwise regressions, an exploratory technique, is consistent with the use of exploratory factoranalysis in our study.


Tests of Hypothesis 1Tax compliance behaviour, the dependent variable, was measured in two differentways: hypothetical-situation compliance and experimentally measured compliance.56

Since our participants provided information about their compliance behaviour attwo different audit rates (0 and 25 percent) for each way of measuring compliance,there were four different variables for compliance behaviour. A regression equationwas estimated for each variable.

Consider first the two equations relating to experimentally measured compli-ance, the results of which are shown in panel A of table 4. Hypothesis 1 was largelyconfirmed. As the theory predicts, variables relating to compliance intention andethics were statistically significant, while the variables relating to attitudes and sub-jective norms were not significant. Furthermore, these variables predicted thedependent variable reasonably well.57 On the other hand, the variables relating tobehavioural control were not significant.

For the two ethics variables, only the P score (moral reasoning) was significant,and it was the most significant variable in the regression equation. For the twointent variables, intent to comply was significant for both audit rates, while intentto overstate deductions was significant only in the 0 percent audit case. This isreasonable, since trying to defend fictitious expenses in an audit is more difficultthan defending concealed income.

Next, consider the equations relating to hypothetical-situation compliance, asshown in panel B of table 4. The results are similar. Again, hypothesis 1 wasgenerally confirmed in that variables relating to compliance intention and ethicswere statistically significant, while the variables relating to attitudes and subjectivenorms were not significant. Behavioural control again was not significant. However,these equations did not predict the dependent variable as well.58 For the two ethicsvariables, it was the variable for anti-establishment views (the A score) that wassignificant.

The above results used the data for 0 percent and 25 percent audit rates inseparate equations. It was also possible to pool the compliance data for these twocompliance rates and thus estimate an equation based on 156 observations (2 foreach participant) rather than 78. The purpose of this procedure was to create anadditional independent variable, the audit rate, which may be thought of as an-other type of perceived behavioural control. The results in panel C of table 4 show

56 Results (not shown) were obtained from a model using “response to cases” (the factor scoreobtained based on the participants’ responses to the four case scenarios presented to them).Consistent with the models using hypothetical compliance measures presented in table 4, theadjusted R square is only 20.1 percent. Only two variables, “monetary considerations” and“intent to comply,” are significant. While “monetary considerations” make the participantsmore non-compliant, “intent to comply” has a negative influence on non-compliance.

57 The adjusted R-squared values of the two models are 42 and 48 percent, respectively.

58 The adjusted R-squared values are substantially less than those for the equations relating toexperimentally measured compliance (21 percent and 25 percent versus 42 percent and 48 percent).


TA

BL

E 4

Ord

inar

y Le

ast

Squ

ares

(O

LS)

Res

ults

from

Mod

els

Usi

ng D

iffe

rent

Com

plia

nce

Mea

sure

s

Pane

l A: O

LS

resu

lts fr

om fu

ll m

odel

rel

atin

g to

exp

erim

enta

lly m

easu

red

com

plia

nce

(N =

78) D

epen

dent

var

iabl

e

Com

plia

nce

at 0

% a

udit

rate

Com

plia

nce

at 2

5% a

udit

rate

bSt

anda

rd e

rror

tSi

gnifi

canc

eb

Stan

dard

err

ort

Sign

ifica

nce

Vari

able

Inte

rcep

t . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .

-0.0

100.

094

-0.1

030.

918

-0.0

350.

117

-0.3

000.

765

P sc

ore

(A)

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .

0.01

60.

003

6.28

00.

000

0.01

70.

003

6.75

80.

000

Inte

nt to

com

ply

(B)

. . .

. . .

. . .

. . .

. . .

. . .

0.29

70.

093

3.19

10.

002

0.27

80.

105

2.65

70.

010

Adv

anta

ge o

f sys

tem

(C)

. . .

. . .

. . .

. . .

. . .

0.06

50.

034

1.92

90.

058

0.16

50.

094

1.74

60.

086

Inte

nt to

ove

rsta

te d

educ

tions

(D)

. . .

. . .

.-0

.041

0.01

8-2

.319

0.02

3T

hird

-par

ty r

epor

ting

(E)

. . .

. . .

. . .

. . .

. .0.

218

0.15

91.

367

0.17

7U

nder

grad

uate

s (F

) . .

. . .

. . .

. . .

. . .

. . .

. .

0.07

30.

086

0.85

40.

396

A ¥

B .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

-0.0

070.

003

-2.5

010.

015

-0.0

050.

003

-1.7

640.

083

E ¥

F .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

0.28

20.

102

2.77

40.

007

A ¥

C .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .-0

.002

0.00

3-0

.594

0.55

5A

¥ E

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

-0.0

110.

004

-2.7

790.

007

B ¥

C .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .-0

.030

0.03

3-0

.932

0.35

5B

¥ E

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

.-0

.291

0.16

3-1

.780

0.08

0C

¥ E

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

0.16

00.

126

1.26

50.

211

A ¥

B ¥

E .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .0.

007

0.00

51.

415

0.16

2A

¥ C

¥ E

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

-0.0

060.

004

-1.6

600.

102

B ¥

C ¥

E .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .0.

143

0.05

22.

769

0.00

7

Adj

uste

d R

-squ

ared

. . .

. . .

. . .

. . .

. . .

. . .

.0.

424

0.48

2M

odel

F (s

igni

fican

ce)

. . .

. . .

. . .

. . .

. . .

.12

.343

(0.0

00)

5.77

7 (0

.000

)

(Tab

le 4

is c

ontin

ued

on th

e ne

xt p

age.

)

56 ■ canadian tax journal / revue fiscale canadienne (2005) vol. 53, no 1T

AB

LE

4Co

ntin

ued

Pane

l B: O

LS

resu

lts fr

om fu

ll m

odel

rel

atin

g to

hyp

othe

tical

-situ

atio

n co

mpl

ianc

e (N

= 7

8)

Dep

ende

nt v

aria

ble

Hyp

othe

tical

com

plia

nce

at 0

% a

udit

Hyp

othe

tical

com

plia

nce

at 2

5% a

udit

bSt

anda

rd e

rror

tSi

gnifi

canc

eb

Stan

dard

err

ort

Sign

ifica

nce

Vari

able

Inte

rcep

t . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .

0.64

80.

066

9.80

10.

000

0.87

10.

052

16.8

220.

000

Inte

nt to

com

ply

. . .

. . .

. . .

. . .

. . .

. . .

. . .

0.17

60.

041

4.26

50.

000

0.12

20.

039

3.15

30.

002

A sc

ore

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .

-0.0

350.

015

-2.3

990.

019

-0.0

350.

012

-2.8

150.

006

Und

ergr

ads

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

.-0

.159

0.10

8-1

.472

0.14

5T

hird

-par

ty r

epor

ting

. . .

. . .

. . .

. . .

. . .

. .

0.06

10.

035

1.76

10.

082

Mon

etar

y co

nsid

erat

ions

. . .

. . .

. . .

. . .

. .

0.05

50.

039

1.42

50.

158

Adj

uste

d R

-squ

ared

. . .

. . .

. . .

. . .

. . .

. . .

.0.

253

0.20

6M

odel

F (s

igni

fican

ce)

. . .

. . .

. . .

. . .

. . .

.9.

695

(0.0

00)

5.99

5 (0

.000

)

(Tab

le 4

is c

oncl

uded

on

the

next

pag

e.)


Pane

l C: O

LS

resu

lts fr

om u

sing

poo

led

data

from

bot

h th

e 0%

and

25%

aud

it ra

tes (

N =

156

) Dep

ende

nt v

aria

ble

Exp

erim

enta

lly m

easu

red

com

plia

nce

Hyp

othe

tical

-situ

atio

n co

mpl

ianc

e

bSt

anda

rd e

rror

tSi

gnifi

canc

eb

Stan

dard

err

ort

Sign

ifica

nce

Vari

able

Inte

rcep

t . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .

0.04

00.

071

0.55

80.

578

1.12

10.

093

12.0

080.

000

0% a

udit

(A)

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

.-0

.065

0.04

5-1

.437

0.15

3-0

.195

0.05

1-3

.833

0.00

0U

nder

grad

uate

s (B

) . .

. . .

. . .

. . .

. . .

. . .

. .

-0.2

870.

095

-3.0

060.

003

Inte

nt to

com

ply

(C)

. . .

. . .

. . .

. . .

. . .

. . .

0.20

50.

066

3.12

20.

002

0.32

10.

097

3.30

00.

001

A sc

ore

(D)

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .-0

.031

0.01

0-3

.241

0.00

1T

hird

-par

ty (E

) .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

0.04

10.

087

0.47

00.

639

Subj

ectiv

e no

rms (

F) .

. . .

. . .

. . .

. . .

. . .

. .-0

.040

0.03

1-1

.289

0.20

0P

scor

e (G

) . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

0.01

60.

002

8.81

20.

000

Inte

nt to

ove

rsta

te d

educ

tions

(H)

. . .

. . .

.-0

.026

0.01

3-2

.051

0.04

2A

dvan

tage

of s

yste

m (I

) . .

. . .

. . .

. . .

. . .

. .

0.06

90.

024

2.88

40.

005

A ¥

C .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .0.

063

0.05

11.

246

0.21

5B

¥ C

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

-0.2

100.

093

-2.2

470.

026

B ¥

E .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

-0.0

450.

097

-0.4

630.

644

C ¥

E .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .-0

.317

0.14

8-2

.148

0.03

3C

¥ F

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

0.00

50.

036

0.14

50.

885

E ¥

F .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

-0.0

710.

038

-1.8

720.

063

C ¥

G .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .-0

.004

0.00

2-2

.187

0.03

0C

¥ H

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

-0.0

270.

013

-2.1

170.

036

B ¥

C ¥

E .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .0.

344

0.14

62.

355

0.02

0B

¥ E

¥ F

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

0.09

30.

045

2.05

40.

042

Adj

uste

d R

-squ

ared

. . .

. . .

. . .

. . .

. . .

. . .

.0.

420

0.33

0M

odel

F (s

igni

fican

ce)

. . .

. . .

. . .

. . .

. . .

.17

.061

(0.0

00)

6.44

3 (0

.000

)

Not

e: S

ee th

e te

xt fo

r th

e de

velo

pmen

t of t

hese

var

iabl

es.

TA

BL

E 4

Conc

lude

d


that the audit rate was not significant in the equation for experimentally measuredcompliance but was significant for hypothetical-situation compliance. Other resultswere generally unchanged, except that ethics was no longer significant in the equa-tion for hypothetical-situation compliance. Thus, the results in panel C of table 4suggest that ethics and attitudinal variables had greater influence on compliancethan did perceived behavioural control, which was affected by the audit rate.

The TPB depicted in figure 1 says that perceived behavioural control, inten-tions, and ethics are generally important in determining compliance. However,simple economic theory suggests that for the particular parameterization of thelaboratory environment (for both hypothetical and experimental environments),perceived behavioural control is likely not relevant (given the penalty for non-compliance, changing audit rates should have no effect on compliance behaviour).59

The regression using the experimentally measured compliance data was consistentwith this amended hypothesis, while the regression using the hypothetical-situationcompliance data was not consistent with it.

Tests of Hypothesis 2Behavioural intention—the dependent variable—was measured in two ways: intentto comply and intent to overstate deductions. In table 5, the equation for the formerdid well in explaining the dependent variable, while the equation for the latter hadrelatively low explanatory power.60 Support for hypothesis 2, which is based on theTPB, was poor: while attitudes were significant in both equations, subjective normswere significant only in the intent-to-comply equation, and neither perceivedbehavioural control nor ethics was significant in either equation. In conclusion,our results appear to support the TPB in a modified form that includes only thethree significant variables.

C O N C L U S I O N

As noted in the introduction to this article, there were two purposes to this study.Both of these purposes related to our use of tax-specific terminology in our in-structions to our participants, which was in contrast to the traditional practice inexperimental economics of using neutral terms such as “fee” for income, “check”for audits, and “shortfall payment” for penalties. The use of tax terminologyallowed for more realistic modelling of the tax setting and the introduction ofpsychological variables into the analysis.

The first purpose was to compare alternative measures of tax compliance behav-iour, which allowed the researcher to directly study how compliance changed witha change in an underlying tax-related parameter such as the audit rate. These twoalternatives involved measures based on hypothetical situations and measures based

59 See the last paragraph of the section headed “Levels of Compliance” above.

60 The adjusted R-squared values are 58 percent and 12 percent, respectively.


TA

BL

E 5

OLS

Res

ults

from

Ful

l Mod

el R

elat

ing

to In

tent

ions

(N

= 7

8)

Dep

ende

nt v

aria

ble

Inte

nt to

com

ply

Inte

nt to

ove

rsta

te d

educ

tions

bSt

anda

rd e

rror

tSi

gnifi

canc

eb

Stan

dard

err

ort

Sign

ifica

nce

Vari

able

Inte

rcep

t . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .

0.16

40.

111

1.47

70.

144

-0.4

100.

213

-1.9

310.

057

Mon

etar

y co

nsid

erat

ions

(A)

. . .

. . .

. . .

. .

-0.4

470.

082

-5.4

520.

000

0.50

50.

214

2.36

20.

021

Adv

anta

ge o

f sys

tem

(B)

. . .

. . .

. . .

. . .

. . .

-0.0

540.

147

-0.3

670.

715

0.53

10.

214

2.48

20.

015

Subj

ectiv

e no

rms (

C)

. . .

. . .

. . .

. . .

. . .

. . .

0.37

60.

080

4.68

80.

000

Dut

ies o

f a c

itize

n (D

) . .

. . .

. . .

. . .

. . .

. . .

0.43

10.

076

5.70

40.

000

A sc

ore

(E)

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .

-0.0

440.

027

-1.6

050.

113

Thi

rd-p

arty

rep

ortin

g (F

) . .

. . .

. . .

. . .

. . .

-0.2

990.

214

-1.3

980.

166

A ¥

C .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .0.

141

0.08

01.

759

0.08

3B

¥ D

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

0.15

90.

088

1.80

80.

075

B ¥

E .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

-0.0

500.

037

-1.3

600.

178

C ¥

D .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. . .

. .-0

.177

0.08

3-2

.142

0.03

6

Adj

uste

d R

-squ

ared

. . .

. . .

. . .

. . .

. . .

. . .

.0.

581

0.12

2M

odel

F (s

igni

fican

ce)

. . .

. . .

. . .

. . .

. . .

.12

.847

(0.0

00)

4.56

4 (0

.005

)

Not

e: S

ee th

e te

xt fo

r th

e de

velo

pmen

t of t

hese

var

iabl

es.


on experimental economics. We tested which measures were best correlated withpast involvement in tax evasion. Our conclusion was that experimentally basedmeasures were superior in that the correlations were higher, were in the expecteddirection, and were statistically significant in all four cases.

The second purpose of this study was to test a particular psychological theory oftax compliance, the TPB of Ajzen,61 in the context of experimental economics. Asapplied to tax compliance, this theory postulates the relationships in figure 1—namely, that attitudes, subjective norms, perceived behavioural control, and ethicsinfluence intention, while intention, perceived behavioural control, and ethics influ-ence behaviour. The statistical analysis in the paper mostly confirmed the theory.The theory was particularly effective in explaining experimentally measured compli-ance behaviour—an important point, since this measure of compliance was foundto have a relatively high correlation with actual tax evasion behaviour on tax returns.

Like other researchers,62 we found that economic variables did not go very farin explaining tax compliance behaviour. In our experiment, the audit and penaltyrates were such that any participant who is maximizing expected take-home paywould report no income for tax purposes; but in fact our participants reported 55to 62 percent of their income. Psychological factors were also found to be impor-tant in the case situations, in which the particular tax context appeared to be quiteimportant. Participants were more inclined to evade taxes when there appeared tobe a genuine reason for such tax evasion, as in the case of the poor student who wasfunding her university education through restaurant tips. Similarly, in the statisticalanalysis of hypothesis 1, it was found that perceived behavioural control variablessuch as third-party reporting, penalties, and audit rates were not as statisticallysignificant as psychological factors such as ethics. Of course, this should not sur-prise tax practitioners, who are very familiar with clients fully complying with thetax laws despite very low audit rates.

As a side issue, we discovered that participants’ average responses to the casesituations suggested that the ranking of the employment situations in which theywere most to least likely to evade tax on income was as follows: restaurant worker(mean 1.34), home handyman (mean 1.24), artist (0.71), and computer consultant(0.43). Again, it appears that the participants were more sympathetic to and moreinclined to behave like the characters in the cases when there appeared to be agenuine reason for such tax evasion.

Our study is subject to a number of limitations. First, it is possible that despiteour clear instructions that the participants’ behaviour in the experiment shouldreflect real-life taxpaying decisions, their tax compliance decisions in the experi-ment did not reflect such decisions. Second, the participants’ admissions of past

61 Supra note 12.

62 Alm and McKee, supra note 6; and James Alm, Isabel Sanchez, and Ana De Juan, “Economicand Noneconomic Factors in Tax Compliance” (1995) vol. 48, no. 1 Kyklos 3-18.


63 Scott J. Boylan and Geoffrey B. Sprinkle, “Experimental Evidence on the Relation BetweenTax Rates and Compliance: The Effect of Earned vs. Endowed Income” (2001) vol. 23, no. 1The Journal of the American Taxation Association 75-90.

64 Note, however, that the tax experiment in Boylan and Sprinkle, ibid., was for a single round.Therefore, it is uncertain whether their results extend to our experimental context, where weadministered the same treatment combination over multiple rounds. Additionally, James Alm,Betty R. Jackson, and Michael McKee, “An Experimental Investigation of Indirect Audit Effectson Individual Compliance Behavior” (manuscript, University of Tennessee, Department ofEconomics, 2004), whose participants earned their income, found results similar to earlierresearch in which participants were endowed with income. The results in Alm, Jackson, andMcKee appear to suggest that it does not matter whether individuals earn their income or areendowed with their income when they make decisions over multiple rounds, as in our study.

evasion may have been influenced by concerns regarding self-presentation or mayhave been affected by poor memory regarding such behaviour. Third, research byBoylan and Sprinkle63 suggests that participants’ reactions when they are endowedwith income is different from their reactions when they are required to earn theirincome, at least in the context of changes in tax rates.64 Further, individuals’compliance behaviour may depend on the source of their income, and individualsmay also differ in the extent to which their non-compliance is intentional orunintentional. Future research can explore these issues in greater detail by varyingthe source of income provided to see whether such differences lead to differencesin compliance behaviour and by differentiating between intentional and uninten-tional compliance behaviour. Fourth, our sample size was modest, especially ofMBA students. Therefore, related results should be interpreted with caution.

We urge future researchers to replicate our study with a larger sample and touse confirmatory statistical techniques such as path analysis or structural equationsto test the robustness of our findings.

Subsequent research to confirm our findings and to determine how best to applythem in policy contexts is needed before such findings can be applied to policydecisions.

Attitudes, Incentives, and Tax Compliance

Documents

Transcript of Attitudes, Incentives, and Tax Compliance